By Zachary Sullivan,2014-08-04 16:44
10 views 0

Committee Examining Radiation Risks of Internal Emitters



    12th Meeting, December 16, 2003 INFO 12-E

    Conference Room 8

    DEFRA, Ashdown House

    123 Victoria Street

    London SW1E 6DE

    SAHSU Report on Bradwell of March 14, 2003

    1. Attached for the Committee’s information is a copy of a the SAHSU RIF unit dated March 2002. This was sent to members by post on September 19 at the request of the CERRIE subcommittee on epidemiology. It is circulated here in order to give it a formal CERRIE INFO number.


    Nov 13




About SAHSU and Limitations of RIF reports

The Small Area Health Statistics Unit (SAHSU)

    SAHSU was established by the Government in response to considerable scientific and public interest in the distribution of diseases across small areas, which arose following the identification of a „cluster‟ of childhood leukaemia near the

    Sellafield nuclear plant in 1983. SAHSU is located in the Department of Epidemiology & Public Health, Imperial College School of Medicine, London.

    SAHSU incorporates national cause-specific data on deaths (from 1981), cancers from the national cancer registry (from 1974), hospital admissions (from 1992) and congenital malformations (from 1983), using the postcode of residence to locate cases to within 100 m. (There are currently around 2 million postcodes in use in the UK.) The system also holds postcoded data on births (from 1981).

    Terms of reference of SAHSU are as follows: to examine quickly and advise on reports of unusual clusters of disease, particularly in the neighbourhood of industrial installations; to carry out detailed epidemiological enquiry of routine health statistics, and, where available, relevant environmental data, particularly in the neighbourhood of industrial installations; in

    collaboration with other scientific groups, to build up background information on the distribution of disease amongst small areas so that specific clusters can be placed in proper context; to explore and develop methods for the study of available statistics in order to detect reliably any unusual incidence of disease; and to develop the methodology for analysing and interpreting statistics relating to small areas.

Reports of Disease Clusters

    This SAHSU report relates the number of cases observed in a specified area to the number expected for a typical population of the same size, age structure and socioeconomic profile as the neighbourhood in question.

    Using an Oracle database and a Geographical Information System (ArcInfo), incident cancer cases, deaths, congenital malformations or hospital admissions occurring among residents living near sources of environmental pollution can be located and rates of disease calculated by linking these events, via the postcode, to the underlying populations at risk. Population data are available for 1981 and 1991 census enumeration districts (around 440 people on average), and 1971 census wards (around 10,000 people). Small area deprivation measures (eg, Carstairs‟ index), which for certain conditions are strongly predictive of mortality and cancer incidence, are also obtained from census statistics and used to adjust disease rates for possible confounding by socio-economic variables.

A description of the SAHSU approach to analysis of data around point sources can be found in Elliott et al, (1 and 2) and

    of the RIF in Aylin et al (3).


    A recent study of cancer mortality around the Bradwell site found excess mortality from breast cancer, prostate cancer and all malignancies in some of the 26 wards surrounding Blackwater Estuary over the years l995-1999.~ These findings were based on very few cases per ward, and were preliminary results that have not yet been published in a peer-reviewed journal. The hypothesis being tested in the study was that ‘risk of cancer mortality is associated with particles cartying ant

    bropogenic radioactivity, including discharges from Bradwell Nuclear Power Station to the Blackwater, which present an inhalation and ingestion hazard since they are subject to resuspension, sea-to-land transfer, and recycling into rivers.’

    SAHSU was asked by the North Essex Health Authority to replicate the Busby et al study, and reported to the Health Authority in July 2001. Discrepancies were noted between the Busby et al and the SAHSU results. In the area covered by


the study, the Post Office terminated all postcodes beginning “CM9 7” in April 1995 (and reallocated the addresses to

    postcodes beginning CM9 4, CM9 5, and CM9 6). These addresses were all in the Heybridge and Maldon areas, which

    were the areas with the largest discrepancies. At the time of the original report, the postcode directories used by SAHSU were only able to provide an estimate for postcode termination dates and did not allocate death records to wards if the postcode was invalid (which includes postcodes that had been terminated more than a year previously). When a postcode is terminated, it does not go out of use immediately; the Post Office will continue to deliver mail with terminated postcodes for the following eighteen months to two years. Terminated postcodes are frequently given by the “informant” when

    registering a death and hence appear in the deaths database at Office for National Statistics (ONS), which is then provided to SAHSU. Since the previous report, SAHSU has gained access to the Gridlink Central Postcode Directory used by ONS, which allows it to allocate terminated postcodes to wards.


    In this study, we have again attempted to recreate the methods described in Busby et al. using the ONS national mortality data held by SAHSU and the new Gridlink Central Postcode Directory supplied by ONS.

    The study area consisted of 26 wards included in the Busby et al study, which surround the Blackwater Estuary, North Essex. Mortality from all cancers, breast cancer and prostate cancer was investigated for the years 1995-1999, using South East England as the comparison population.

    The following definitions of conditions were used: all malignant neoplasms (ICD9 140-208), female breast cancer (ICD9 174), and prostate cancer (ICD9 185). Standardised (by age, sex and Carstairs‟ deprivation score). Mortality Ratios were

    calculated, using England and Wales as the standard population.

Further analysis used Bayesian disease-mapping techniques to stabilise risk estimates based on small numbers at ward 5. The key idea of this approach is to “smooth” the estimates in each ward by combining information about the ward-level

    specific disease rate with information about the disease rates in other wards in the study region. The resulting relative risk estimate for each ward is a form of weighted average of the observed risk and the mean relative risk in the remaining wards; the amount by which each SMR is weighted or smoothed is inversely proportional to the expected count in that ward.


    Breast cancer (table 2): there was no significant excess risk of mortality from breast cancer in any ward, and no overall statistically significant excess risk

    (SMR=~0.95 (95%CI 0.77-1.15)).

    Prostate cancer (table 3): there was no significant excess risk of mortality from prostate cancer in any ward, and no overall statistically significant excess risk

    (SMR=l.04 (CI 0.83-1.28)).

    All malignancies (table 4): there were nominally significant excess relative risks at the 95% level in two wards, although overall there was no significant excess risk of mortality from all malignancies in the study region. It should be noted that since a large number of tests have been undertaken (for both specific cancer sites and all malignancies combined in each of 26 wards) the finding of two nominally significant excess risks should not be considered unusual. The SMR for all malignancies for the whole study region combined was (SMR males = 1.01 (CI 0.93-1.1); females = 0.98 (CI 0.9-1.07)).






    It should be noted that SAHSU RIF reports are subject to a number of limitations. They will not account for errors in data supplied to SAHSU , which may include under-registration, duplicate registration, errors in coding and classification of cancer cases, death certification errors, biases or errors in hospital admissions data and errors in postcoding and geographic linkages. Factors affecting hospital admissions including hospital and primary care effects will not have been taken into account. In addition, problems associated with cluster investigation may affect SAHSU reports. These include difficulties in defining boundaries, which could affect the observed rates, the risk of multiple comparisons and small numbers of events. SAHSU RIF reports therefore need to be interpreted with caution and with expert local knowledge.

    The RIF reports, which are made available to local health authorities to inform their own investigations, are not themselves published as SAHSU studies, but are intended to provide health authorities with a rapid initial screen of the health statistics

    related to a particular point source. They use the postcode to give information on the geographic location of cases and links to the underlying socio-demographic population statistics (which allows, for example, analyses to be carried out based on distance from a point source). Given the rapid turnaround of a few working days, the RIF reports are necessarily based only on data held routinely on the database, without any scope for further checking of the data. In SAHSU‟s national studies,

    since overall a very high proportion of cases (for mortality, 98.9%) have a valid postcode that can be geographically referenced, these studies are far less affected by local postcoding anomalies than potentially is the case around a single point source in a specific locality.


    The numbers in SAHSU RIF reports are not derived directly from the ward codes, but through the postcode on each record using various electronic postcode directories. The principal differences between numbers of cancers in the original RIF report provided to the Health Authority and numbers subsequently provided by ONS are concentrated in just five of the 26 wards: Heybridge East, 1-leybridge West, Maldon East, Maldon North West, and Maldon South. As noted, these are the areas where postcodes were terminated and reallocated. For all malignancies, these differences total 53 deaths for males and 48 for females.

Checks have shown that over the country as a whole, the SAJ-ISU RIF allocates 98.9% of all death records, and 99.3% of

    death records with a valid postcode, to a ward. On average, approximately 50,000 postcodes, around 3% of the UK total of 1.6 million, are terminated each year, although the numbers fluctuate greatly from year to year. As noted above, ONS has now supplied SAI{SU with the Gridlink Central Postcode Directory (CPD) which contains terminated postcodes. This revised SAHSU RIF report, using the ONS CPD, shows numbers of deaths that are very close to the numbers in ONS published data.

Table 5. Comparisons of numbers of cancer deaths* in 26 wards in Essex,

    1995-1999: ONS, SAHSU.

Cancer site(s) Sex ONS SAHSU

     No. Diff

    All Malingnancies+ M 588 591 +3

     F 532 537 ?5

     Total 1,120 1,128 ?8

    Breast cancer F 104 104 0

    Prostate cancer M 84 85 ?1

    * The ONS figures relate to deaths for which the registration was made in 1995-1999. The SAHSU figures relate to deaths which actually occurred in 1995-1999; on this basis, the ONS figures would have been higher by 2 for all cancers, and lower by 1 for prostate cancer.

    + All malignant neoplasms, ie codes 140-208 inclusive in the Ninth Revision of the WHO Classification of Diseases (WHO: Geneva, 1975).



1 Elliott P, Westlake AJ, Kleinschmidt I, Hills M, Rodrigues L, McGales P, Marshall K, Rose G. The Small Area Health

    Statistics Unit: a national facility for investigating health around point sources of environmental pollution in the United Kingdom. J Epidemiol Community Health 1 992;46:345-349.

    2 Elliott P, Kleinschmidt I, Westlake AJ. Use of routine data in studies of point sources of environmental pollution. In: Elliott P, Cuzick J, English D, Stem R, eds. Geographical and Environmental Epidemiology: methods for Small-Area

    Studies. Oxford: Oxford University Press, 1992;106-1

    3. Aylin P, Maheswaran R, Wakefield J, Cockings 5, Jarup L, Arnold R, Wheeler G, Elliott P. A national facility for small area disease mapping and rapid initial assessment of apparent disease clusters around a point source: the UK Small Area Health Statistics Unit. J Public Health Med 1999 21:289-298

4. Busby C, Dorfman P, Bramhall R. Cancer Mortality and proximity to Bradwell Nuclear Power Station in Essex, 1995-

    1999. Preliminary results. Green Audit: Aberystwyth. March 2001.

    5. Wakefield JC, Best NG, Waller L. Bayesian approaches to disease mapping. In: Elliott P, Wakefield JC, Best NG, et al., editors. Spatial epidemiology: methods and applications, Oxford: Oxford University Press, 2000. 10427.



Report this document

For any questions or suggestions please email